Bars are expected to play an important role in galactic dynamics. Since the rest-frame wavelength of near-infrared images of high redshift galaxies lies in the optical, a reliable comparison of low and high redshift bars requires an understanding of how the properties of a bar depend on the observing wavelength. For this purpose, we compare bar strengths for galaxies from the OSUBSGS sample in the B- and H-bands, using the gravitational bar torque method. We find that the ratio of bar torques in the B- and H-bands is fairly close to unity (QB/H ˜ 1.25) and independent of the morphological type. Hence, we can make a good estimate of the ``real'' bar strength for galaxies at high redshift from the observed one, even if the morphological classification were uncertain. Also, we show that the gravitational bar torque method using a polar grid gives satisfactory results when applied to degraded images that mimic high redshift images. Typical errors in bar strength remain below 10% as long as the total bar length is at least ˜10 pixels.